Multiple cluster structures and mixture models: recent developments for multilevel data

This paper deals with the problem of identifying different partitions of a given set of units obtained according to different subsets of the observed variables (multiple cluster structures). Procedures have been recently developed for detecting multiple cluster structures in a data matrix. In a previous paper we proposed a strategy which rely on model-based clustering methods and on a comparison between mixture models using model selection criteria. A generalization of this method which allows the analysis of data matrices with nested data structures is considered. The usefulness of the new method is shown using simulated and real examples.